Just this skill, with its reference files, at the pinned commit:
Or the whole set. The skills link to each other's format files, so the set works best together:
No install: tell your agent
SKILL.md
Open .mdLaunch
A launch on Solana is coupled to a transaction that either lands or does not. This skill writes a runbook in which nothing public happens until that transaction is finalized and checked, so a slip moves the schedule instead of breaking it. It is a sequence, not a list of ideas.
It does not deploy anything (ai-kit’s
/deploy
does), and it does not answer legal questions
(crypto-legal-skill does).
Context handoff
At start, read .claude/context/idea.md, build.md, positioning.md and pricing.md if
present. Take the one-liner and the words to avoid from positioning.md, and what works
today from build.md. If .claude/context/launch.md exists, resume it: report its status
and the next open gate instead of starting over.
Writes .claude/context/launch.md in the format defined in
runbook.md, and the drafts it lists under
launch/ in the project.
Mode 1: plan (any time before T-24h)
- Pin the event. Ask only what the context files do not answer:
- which onchain event: program deploy or upgrade, mint or TGE, liquidity add, airdrop claim, audit release (several are fine; each gets its own G0 row)
- planned T-0 as a date in UTC, and who signs (deployer key or multisig, threshold)
- whether a token is involved, and in which jurisdictions users are
- partners who will amplify, and the team’s own channels
- Build the gate list from runbook.md. Keep the rows that apply, compute T- dates from T-0, and name an owner per row. Every gate gets a condition that is true or false.
- Legal gates L1 (T-30d) and L2 (T-7d) are hard gates. Tell the user to run
crypto-legal-skill’s
/launch-checklist, and record counsel’s sign-off in the runbook with name and date. That skill writes no file, so the runbook is where the sign-off lives. Do not mark either gate passed on the user’s say-so that “legal is fine”. - Readiness is engineering’s. If the verifiable build, IDL or audit is not done, route
to ai-kit (
/build-program,/deploy,/audit-solana) and leave the T-14d gate open. - Fill the key-holder table and the abort criteria from slip-and-abort.md. Mark every step reversible or not. Two names per key. Decide the abort criteria now, not on the day.
- Plan the channels with channels.md: the order, the partner kit, the Superteam chapter and other ecosystem accounts, and a separate day for Product Hunt or Show HN.
- Draft the artifacts in artifacts.md, including the hold, slip and abort messages. Pages start from launch-post.html and stay one self-contained HTML file. Use writer-style-skill and content-gen-skill for voice and long-form if present.
- Write
.claude/context/launch.mdwith statusplanning, thenarmedonce every T-24h gate has passed.
Mode 2: execute (launch day)
Run /launch again on the day. Walk the gates in order and write a UTC timestamp into the
runbook as each one passes.
- Before G0, nothing public. No post, no scheduled post, no partner “go”, no Product Hunt.
- G0 is a finalized transaction checked from a second RPC, against the event’s row in runbook.md. A wallet’s success toast is not G0.
- After G0, the L+ clock runs. Fill only the address placeholders, then release the artifacts in channel order.
- On a slip: send the partner hold message at once, set a decision time, and post the public slip message only if a time was public. See slip-and-abort.md.
- On an abort: run the abort steps there. If part of the onchain state already landed, this is now an incident; use incident-comms.
- At L+24h: set status
landed, log what slipped and why, and list the listings still pending.
Rules
- No public copy carries a wall-clock time. Partners post after your post.
- The ticker and the mint address are published together, after G0, from your account first.
- Revoking authorities, adding first liquidity and opening claims are separate steps with their own gate, never folded into the launch batch to save time.
- Anything legal or tax-related goes to crypto-legal-skill. Say so and stop. Do not improvise an answer.
One install for the whole kit
Superteam Brasil builds these skills as part of the Solana AI Kit, its open-source Claude Code and Codex kit for Solana. 15 specialist agents, 32 commands, 4 MCP servers and 46 skill packs, behind a firewall that keeps agents away from wallet keypairs.
- Specialist agents
- 15
- Commands
- 32
- Skill packs
- 46
- MCP servers
- 4